V2:

About this version

Exchange Q group: 950351015, welcome to play and share with each other.

Telegram: https://t.me/+nbU3j7rLEYZkNzVl

The polka dots I made before were always badly done and polluted. After many, many attempts, made new ones. It's so much better than the previous one, easier to trigger and less contamination.

Difference Extraction Method 3.0 + real life production.

Right to try more and more, depending on the base film and the effect you want to achieve.

Suggest shoes with black footwear or Loafers, leather boots, etc., with the feeling of better look.

In order to facilitate the use of the future I produced LORA trigger words are all tutututu, four tu

This model can be written as: (tutututu),pantyhose.

Why this LORA?

Although there are a lot of similar LORA, I found that on the one hand, most of these LORA are not very close to reality, many of the current fashionable clothing is not, and not realistic enough.

On the other hand, the quality of LORA is uneven, and many LORA have a big impact on the character itself and the picture, which is not convenient to use. So I intend to do a series of clothing LORA.

Usage:

1, to ensure that you correctly use LORA

2, CLIP terminate the number of layers is recommended to be set to 2

3, the weight to open between 0.4-1, you can try more.

4, the trigger word is: (tutututu), pantyhose, (can also be written as tutututu, polka dot pantyhose, trigger the wave point is bigger, but polka dot the word will be more likely to have an impact on other aspects of the image, you can also add BREAK)

5、Negative do not add too much, otherwise it will have an effect on the effect of LORA

6, cue word guidance coefficient (CFG Scale) can start from 7 to try!

If you encounter problems in use, you can leave me a message!

Looking forward to your work!

Looking forward to your work!

Looking forward to your work!

Statement:

1、You should take full responsibility for any creative works using this model.

2、You must not use this model to intentionally create or share illegal or harmful output or content, and avoid using this model for malicious, hurtful, defamatory, fraudulent or political purposes.

3、Please notify me if you use this model for commercial purposes, thank you.

Frequently Asked Questions:

1, What is the difference extraction method?

As the name suggests, it is simply to extract the differences to make LORA, our general LORA is to take the picture directly training, so the trained LORA will inevitably be contaminated by the bottom of the film as well as other environmental factors, so in the time of the picture will have an unexpected impact on the picture. The LORA produced by the difference extraction method will remove these contaminants, so that the LORA produced even if the weight is relatively high on the picture of the contamination is also very small, the model quality will be better. The difference extraction method is more complicated and requires more steps.

2、Why make this LORA?

Although there are a lot of similar LORA, I found that on one hand, most of this LORA is not very close to reality, many of the more popular clothing is not, and not realistic enough.

On the other hand, the quality of LORA is not uniform, and many LORA have a big impact on the characters and the image, which is not convenient to use. So I intend to do a series of clothing LORA.

3、Why can't I produce the same photos as you?

Everyone's computer configuration and operating environment are different, the use of LORA and plug-in version, plug-in configuration, etc. will also be different, so please do not pursue the exact same photo, create your own perfect work!

4、How to get in touch?

If there is a demand for custom LORA, or other cooperation intentions, you can directly leave a message here, or contact Q331506796 (specify the intention)

5、What have you done to optimise the difference extraction method? Or just casually?

A: I have done a lot of optimisation on the code involved:

1, automated model exploration: automatically generates multiple fusion models from a set of models, similar to the automated process of parameter tuning or model selection, but focused on model fusion.

2. variability analysis based on Euclidean distance: used to compare the difference in weights between two LoRA (Low-RankAdaptation) models.

In variability analysis, Euclidean distance is used to quantify the difference between the weights of two models. Specifically:

Weight representation: the weight of each LoRA model can be regarded as a point in a high-dimensional space. Each element of the weight corresponds to a dimension in the space.

Calculating the difference: by calculating the Euclidean distance between the corresponding weights of two models, we can obtain a value that represents the difference in the relative positions of these two weights in the high-dimensional space.

Interpreting the discrepancy: a larger Euclidean distance means that the two models are more different in that weight; a smaller distance means that they are more similar.

This variability analysis is used to guide decisions on model fusion. For example, if the two models are very different on a particular weight, that weight may need to be treated specifically to ensure that the fused model effectively combines the properties of the two original models.

These optimisations are done entirely for my own material, which has some peculiarities, so it's not a difference extraction in the traditional sense, and there are some trade-offs to be made.

Have fun with it.

交流Q群：950351015，欢迎来玩，互通有无。

Telegram: https://t.me/+nbU3j7rLEYZkNzVl

之前做的波点总是做不好，污染很大。经过很多很多次尝试，做了新的。比之前的好太多了，更容易触发，污染更少。

差异提取法3.0+实拍制作。

权重要多多尝试，根据底膜和想要实现的效果来。

建议鞋子用black footwear或者乐福鞋、皮靴等，搭配感觉更好看。

为了方便使用，以后我制作的LORA触发词全部都是tutututu,四个tu

这个模型可以写成：(tutututu),pantyhose,

为什么要做这个LORA?

虽然类似的LORA有很多，但是我发现一方面这种LORA大多数不是很贴近现实，很多当下比较流行的服装并没有，且不够写实。

另外就是质量方面残次不齐，很多LORA对人物本身以及画面的影响比较大,使用起来不够方便。所以打算做一系列的服装LORA。

使用方法：

1、确保你正确的使用LORA

2、CLIP终止层数建议设置为2

3、权重开0.4-1之间，可以多多尝试。

4、触发词为：(tutututu),pantyhose,（也可以写成tutututu, polka dot pantyhose,触发的波点更大，但是polka dot这个词会更容易对图像的其他方面造成影响，也可以加BREAK）

5、负面不要加的太多，否则会对LORA的效果产生影响

6、提示词引导系数(CFG Scale)可以从7开始尝试

如果您在使用中遇到问题，可以给我留言

期待您的作品！

期待您的作品！

期待您的作品！

声明：

1、您应对使用此模型的任何创意作品承担全部责任。

2、您不能使用该模型故意制作或分享非法或有害的输出或内容，避免将此模型用于恶意、伤害、诽谤、诈骗或政治用途。

3、如果将本模型用于商业用途，请通知我，谢谢。

常见问题：

1、什么是差异提取法？

顾名思义，简单来说就是将差异提取出来制作LORA。我们一般的LORA都是拿图直接训练，这样训练出的LORA不可避免的会受到底膜以及其他环境因素污染，所以在出图的时候会对画面产生预料之外的影响。差异提取法制作的LORA会将这些污染去除掉，这样制作出的LORA权重即使开的比较高对画面的污染也很小，模型质量会更好。差异提取法制作起来更复杂，需要的步骤更多。

2、为什么要做这个LORA?

虽然类似的LORA有很多，但是我发现一方面这种LORA大多数不是很贴近现实，很多当下比较流行的服装并没有，而且不够写实。

另外就是质量方面残次不齐，很多LORA对人物本身以及画面的影响比较大,使用起来不够方便。所以打算做一系列的服装LORA。

3、为什么我出不了和你一样的照片？

每个人的电脑配置以及操作环境都不一样，使用的LORA以及插件版本、插件配置等也会有所区别，所以请不要追求一模一样的照片，创作您自己的完美作品吧！

4、如何取得联系？

如果有定制LORA的需求，或者有其他合作意向，可以直接在这里留言，或者联系Q331506796（注明来意）

5、你对差异提取法做了哪些优化？还是只是随便说说？

答：我对涉及代码进行了大量的优化：

1、自动化模型探索：自动从一组模型中生成多种融合模型，类似于参数调优或模型选择的自动化过程，但专注于模型融合。

2、基于欧氏距离的差异性分析：用于比较两个LoRA（Low-RankAdaptation）模型的权重差异。

在差异性分析中，欧氏距离用于量化两个模型权重之间的差异。具体来说：

权重表示：每个LoRA模型的权重可以看作是高维空间中的一个点。权重的每个元素对应空间中的一个维度。

计算差异：通过计算两个模型相应权重之间的欧氏距离，我们可以得到一个数值，表示这两个权重在高维空间中的相对位置差异。

解释差异：较大的欧氏距离意味着两个模型在该权重上的差异较大；较小的距离则意味着它们较为相似。

这种差异性分析用于指导模型融合的决策。例如，如果两个模型在某个权重上的差异很大，可能就需要特别处理这个权重，以确保融合后的模型能够有效地结合两个原始模型的特性。

这些优化完全是针对我自己的素材进行的，因为我的素材有一些特殊性，所以不是传统意义上的差异提取，要有一些取舍。

祝您玩的开心。

V1:

交流Q群：950351015，欢迎来玩，互通有无。

品牌：某觅

所属系列：限量款波卡圆点

厚度：5D

颜色：black

简介：

这是一个黑色超薄连裤袜LORA，可以改善图中黑丝的质感。

本LORA的特点如下：

1、使用差异提取法精心制作。1024*1024尺寸+bf16，细节拉满。

2、全部采用真人照片高清相机拍摄制作，确保效果真实。

使用方法：

这个LORA在2.5D底膜下出图效果最佳，其次是真人大模型，二次元也能出，但是要仔细微调。

1、确保正确的加载LORA

2、正确的使用触发词tutubd,black pantyhose, 可以给tutubd加点权重，比如：(tutubd),black pantyhose,

3、CLIP 终止层数建议设置为2

4、这个LORA的权重要在0.5-1之间才会有更好的效果。可以从0.6开始尝试。

其他参数根据习惯进行设置即可，

如果您在使用中遇到问题，可以给我留言。

期待您的作品！

期待您的作品！

期待您的作品！

声明：

1、您应对使用此模型的任何创意作品承担全部责任。

2、您不能使用该模型故意制作或分享非法或有害的输出或内容，避免将此模型用于恶意、伤害、诽谤、诈骗或政治用途。

3、禁止将本模型用于商业用途，谢谢。

常见问题：

1、为什么我出不了和你一样的照片？

每个人的电脑配置以及操作环境都不一样，使用的LORA以及插件版本、插件配置等也会有所区别，所以请不要追求一模一样的照片，创作您自己的完美作品吧！

2、权重调高以后图片出现大面积马赛克怎么办？

尝试调低权重，或者降低其他LORA的权重，或者打开LoRA Block Weight插件。

3、如何取得联系？

如果有定制LORA的需求，或者有其他合作意向，可以直接在这里留言，或者联系Q331506796（注明来意）

推荐模型：

会魔法de大麻瓜：

【麻瓜lora】日料店、【麻瓜lora】扶梯视角、【麻瓜lora】夜店蹦迪酒吧，一起来喝酒、【麻瓜lora】副驾驶女友、【麻瓜lora】落地窗/法式大窗户、【麻瓜lora】真实试衣间自拍

墨幽：

墨幽人造人、墨幽FACE_realistic 、全网首发︱墨幽2.5D_MIX_MoYou2.5D

LEOSAM：

LEOSAM's Instant photo 拍立得/Polaroid 、 LEOSAM's FilmGirl 胶片风 LoRA--全新VELVIA 、 LEOSAM's Clothing +/- Adjuster 服装增减

Brand: a search

The series: limited edition Polka dots

Thickness: 5D

color: black

Introduction:

This is a black ultra-thin pantyhose LORA, which can improve the texture of the black silk in the picture.

The characteristics of this LORA are as follows:

1. Elaborately crafted using differential extraction. 1024*1024 size + bf16, full of details.

2. All photos are taken with real-life high-definition cameras to ensure real effects.

This LORA works best under the 2.5D base film, followed by the real-life large model, which can also be produced in the second dimension, but it needs to be carefully fine-tuned.

1. Make sure to load LORA correctly

2. Correctly use the trigger words tutubd, black pantyhose, you can add some weight to tutubd, for example: (tutubd), black pantyhose,

3. It is recommended to set the number of CLIP termination layers to 2

4. The weight of this LORA should be between 0.5-1 to have a better effect. You can try it from 0.6.

Other parameters can be set according to custom,

If you encounter problems in use, you can leave me a message.

Looking forward to your work!

Looking forward to your work!

Looking forward to your work!

statement:

1. You are solely responsible for any creative work using this mockup.

2. You cannot use this model to intentionally produce or share illegal or harmful output or content, and avoid using this model for malicious, hurtful, defamatory, fraudulent or political purposes.

3. It is forbidden to use this model for commercial purposes, thank you.

common problem:

1. Why can't I post the same photos as you?

Everyone's computer configuration and operating environment are different, and the LORA used, plug-in version, plug-in configuration, etc. will also be different, so please don't pursue exactly the same photos, create your own perfect work!

2. What should I do if there is a large area of mosaic in the picture after the weight is increased?

Try to lower the weight, or lower the weight of other LORA, or turn on the LoRA Block Weight plugin.

3. How to get in touch?

If you have a need to customize LORA, or have other cooperation intentions, you can leave a message here directly, or contact Q331506796 (indicate your intention)

Recommended model:

[Muggle lora] Japanese grocery store, [Muggle lora] escalator perspective, [Muggle lora] nightclub disco bar, let's drink together, [Muggle lora] co-pilot girlfriend, [Muggle lora] floor-to-ceiling windows/large French windows , 【Muggle lora】Real fitting room selfie

Moyou:

Moyou artificial human, Moyou FACE_realistic, the whole network debut︱Moyou 2.5D_MIX_MoYou2.5D

LEOSAM:

LEOSAM's Instant photo Polaroid/Polaroid, LEOSAM's FilmGirl film style LoRA--new VELVIA, LEOSAM's Clothing +/- Adjuster clothing increase or decrease

ブランド: 検索

シリーズ：限定版ポルカドット

厚さ: 5D

色: 黒

導入：

これは、写真の黒いシルクの質感を向上させることができる黒の極薄パンスト LORA です。

このLORAの特徴は以下の通りです。

1. 差分抽出を使用して精巧に作られています。 1024*1024サイズ+bf16、詳細が満載。

2.すべての写真は本物の効果を保証するために、実際の高解像度カメラで撮影されています。

説明書：

この LORA は 2.5D ベースフィルムで最もよく機能し、次に現実の大型モデルが最適です。これは 2 次元でも作成できますが、慎重に微調整する必要があります。

1. LORA が正しくロードされていることを確認してください

2. トリガーワード tutubd、黒パンストを正しく使用します。たとえば、(tutubd)、黒パンスト、などのように、tutubd に重みを追加できます。

3. CLIP 終端層の数を 2 に設定することをお勧めします。

4. より良い効果を得るには、この LORA の重みは 0.5 ～ 1 の間である必要があります。 0.6からお試しいただけます。

他のパラメータはカスタムに従って設定できます。

使用中に問題が発生した場合は、メッセージを残してください。

あなたの仕事を楽しみにしています！

あなたの仕事を楽しみにしています！

あなたの仕事を楽しみにしています！

声明：

1. このモックアップを使用したクリエイティブな作業については、お客様が単独で責任を負います。

2. このモデルを使用して、違法または有害な出力やコンテンツを意図的に作成または共有することはできません。また、このモデルを悪意のある、有害な、中傷的な、詐欺的、または政治的な目的で使用することは避けてください。

3. このモデルを商業目的で使用することは禁止されています。よろしくお願いします。

よくある問題:

1. なぜ私はあなたと同じ写真を投稿できないのですか?

パソコンの構成や動作環境は人それぞれ異なりますし、使用しているLORAやプラグインのバージョン、プラグインの構成なども異なりますので、全く同じ写真を追い求めるのではなく、自分だけの完璧な作品を作り上げてください。

2. ウェイトを増やした後、写真に広い範囲のモザイクが入った場合はどうすればよいですか?

重みを下げるか、他の LORA の重みを下げるか、LoRA Block Weight プラグインをオンにしてみてください。

3. 連絡方法は?

LORA をカスタマイズする必要がある場合、またはその他の協力の意図がある場合は、ここに直接メッセージを残すか、Q331506796 に連絡してください (意図を示してください)

Files with SHA256: 479e3bbb11c9a5c35efc639451076f5448c855c3feb3201053d1f796832918a9